src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Coulomb
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 102     real             *vftab;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 121     charge           = mdatoms->chargeA;
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     vftab            = kernel_data->table_vdw->data;
 127     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[0];
 131     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 132     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 133     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 134     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136     /* Avoid stupid compiler warnings */
 137     jnrA = jnrB = 0;
 138     j_coord_offsetA = 0;
 139     j_coord_offsetB = 0;
 140
 141     outeriter        = 0;
 142     inneriter        = 0;
 143
 144     /* Start outer loop over neighborlists */
 145     for(iidx=0; iidx<nri; iidx++)
 146     {
 147         /* Load shift vector for this list */
 148         i_shift_offset   = DIM*shiftidx[iidx];
 149
 150         /* Load limits for loop over neighbors */
 151         j_index_start    = jindex[iidx];
 152         j_index_end      = jindex[iidx+1];
 153
 154         /* Get outer coordinate index */
 155         inr              = iinr[iidx];
 156         i_coord_offset   = DIM*inr;
 157
 158         /* Load i particle coords and add shift vector */
 159         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 160                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 161
 162         fix0             = _fjsp_setzero_v2r8();
 163         fiy0             = _fjsp_setzero_v2r8();
 164         fiz0             = _fjsp_setzero_v2r8();
 165         fix1             = _fjsp_setzero_v2r8();
 166         fiy1             = _fjsp_setzero_v2r8();
 167         fiz1             = _fjsp_setzero_v2r8();
 168         fix2             = _fjsp_setzero_v2r8();
 169         fiy2             = _fjsp_setzero_v2r8();
 170         fiz2             = _fjsp_setzero_v2r8();
 171         fix3             = _fjsp_setzero_v2r8();
 172         fiy3             = _fjsp_setzero_v2r8();
 173         fiz3             = _fjsp_setzero_v2r8();
 174
 175         /* Reset potential sums */
 176         velecsum         = _fjsp_setzero_v2r8();
 177         vvdwsum          = _fjsp_setzero_v2r8();
 178
 179         /* Start inner kernel loop */
 180         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 181         {
 182
 183             /* Get j neighbor index, and coordinate index */
 184             jnrA             = jjnr[jidx];
 185             jnrB             = jjnr[jidx+1];
 186             j_coord_offsetA  = DIM*jnrA;
 187             j_coord_offsetB  = DIM*jnrB;
 188
 189             /* load j atom coordinates */
 190             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 191                                               &jx0,&jy0,&jz0);
 192
 193             /* Calculate displacement vector */
 194             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 195             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 196             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 197             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 198             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 199             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 200             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 201             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 202             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 203             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 204             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 205             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 206
 207             /* Calculate squared distance and things based on it */
 208             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 209             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 210             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 211             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 212
 213             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 214             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 215             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 216             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 217
 218             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 219             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 220             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 221
 222             /* Load parameters for j particles */
 223             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 224             vdwjidx0A        = 2*vdwtype[jnrA+0];
 225             vdwjidx0B        = 2*vdwtype[jnrB+0];
 226
 227             fjx0             = _fjsp_setzero_v2r8();
 228             fjy0             = _fjsp_setzero_v2r8();
 229             fjz0             = _fjsp_setzero_v2r8();
 230
 231             /**************************
 232              * CALCULATE INTERACTIONS *
 233              **************************/
 234
 235             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 236
 237             /* Compute parameters for interactions between i and j atoms */
 238             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 239                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 240
 241             /* Calculate table index by multiplying r with table scale and truncate to integer */
 242             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 243             itab_tmp         = _fjsp_dtox_v2r8(rt);
 244             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 245             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 246             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 247
 248             vfconv.i[0]     *= 8;
 249             vfconv.i[1]     *= 8;
 250
 251             /* CUBIC SPLINE TABLE DISPERSION */
 252             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 253             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 254             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 255             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 256             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 257             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 258             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 259             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 260             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 261             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 262             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 263
 264             /* CUBIC SPLINE TABLE REPULSION */
 265             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 266             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 267             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 268             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 269             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 270             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 271             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 272             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 273             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 274             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 275             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 276             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 277             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 278
 279             /* Update potential sum for this i atom from the interaction with this j atom. */
 280             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 281
 282             fscal            = fvdw;
 283
 284             /* Update vectorial force */
 285             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 286             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 287             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 288
 289             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 290             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 291             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 292
 293             /**************************
 294              * CALCULATE INTERACTIONS *
 295              **************************/
 296
 297             /* Compute parameters for interactions between i and j atoms */
 298             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 299
 300             /* COULOMB ELECTROSTATICS */
 301             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 302             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 303
 304             /* Update potential sum for this i atom from the interaction with this j atom. */
 305             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 306
 307             fscal            = felec;
 308
 309             /* Update vectorial force */
 310             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 311             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 312             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 313
 314             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 315             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 316             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 317
 318             /**************************
 319              * CALCULATE INTERACTIONS *
 320              **************************/
 321
 322             /* Compute parameters for interactions between i and j atoms */
 323             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 324
 325             /* COULOMB ELECTROSTATICS */
 326             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 327             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 328
 329             /* Update potential sum for this i atom from the interaction with this j atom. */
 330             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 331
 332             fscal            = felec;
 333
 334             /* Update vectorial force */
 335             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 336             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 337             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 338
 339             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 340             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 341             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 342
 343             /**************************
 344              * CALCULATE INTERACTIONS *
 345              **************************/
 346
 347             /* Compute parameters for interactions between i and j atoms */
 348             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 349
 350             /* COULOMB ELECTROSTATICS */
 351             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 352             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 353
 354             /* Update potential sum for this i atom from the interaction with this j atom. */
 355             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 356
 357             fscal            = felec;
 358
 359             /* Update vectorial force */
 360             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 361             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 362             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 363
 364             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 365             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 366             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 367
 368             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 369
 370             /* Inner loop uses 155 flops */
 371         }
 372
 373         if(jidx<j_index_end)
 374         {
 375
 376             jnrA             = jjnr[jidx];
 377             j_coord_offsetA  = DIM*jnrA;
 378
 379             /* load j atom coordinates */
 380             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 381                                               &jx0,&jy0,&jz0);
 382
 383             /* Calculate displacement vector */
 384             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 385             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 386             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 387             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 388             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 389             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 390             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 391             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 392             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 393             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 394             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 395             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 396
 397             /* Calculate squared distance and things based on it */
 398             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 399             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 400             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 401             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 402
 403             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 404             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 405             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 406             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 407
 408             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 409             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 410             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 411
 412             /* Load parameters for j particles */
 413             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 414             vdwjidx0A        = 2*vdwtype[jnrA+0];
 415
 416             fjx0             = _fjsp_setzero_v2r8();
 417             fjy0             = _fjsp_setzero_v2r8();
 418             fjz0             = _fjsp_setzero_v2r8();
 419
 420             /**************************
 421              * CALCULATE INTERACTIONS *
 422              **************************/
 423
 424             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 425
 426             /* Compute parameters for interactions between i and j atoms */
 427             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 428
 429             /* Calculate table index by multiplying r with table scale and truncate to integer */
 430             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 431             itab_tmp         = _fjsp_dtox_v2r8(rt);
 432             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 433             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 434             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 435
 436             vfconv.i[0]     *= 8;
 437             vfconv.i[1]     *= 8;
 438
 439             /* CUBIC SPLINE TABLE DISPERSION */
 440             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 441             F                = _fjsp_setzero_v2r8();
 442             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 443             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 444             H                = _fjsp_setzero_v2r8();
 445             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 446             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 447             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 448             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 449             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 450             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 451
 452             /* CUBIC SPLINE TABLE REPULSION */
 453             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 454             F                = _fjsp_setzero_v2r8();
 455             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 456             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 457             H                = _fjsp_setzero_v2r8();
 458             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 459             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 460             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 461             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 462             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 463             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 464             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 465             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 466
 467             /* Update potential sum for this i atom from the interaction with this j atom. */
 468             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 469             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 470
 471             fscal            = fvdw;
 472
 473             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 474
 475             /* Update vectorial force */
 476             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 477             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 478             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 479
 480             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 481             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 482             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             /* Compute parameters for interactions between i and j atoms */
 489             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 490
 491             /* COULOMB ELECTROSTATICS */
 492             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 493             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 494
 495             /* Update potential sum for this i atom from the interaction with this j atom. */
 496             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 497             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 498
 499             fscal            = felec;
 500
 501             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 502
 503             /* Update vectorial force */
 504             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 505             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 506             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 507
 508             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 509             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 510             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 511
 512             /**************************
 513              * CALCULATE INTERACTIONS *
 514              **************************/
 515
 516             /* Compute parameters for interactions between i and j atoms */
 517             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 518
 519             /* COULOMB ELECTROSTATICS */
 520             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 521             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 522
 523             /* Update potential sum for this i atom from the interaction with this j atom. */
 524             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 525             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 526
 527             fscal            = felec;
 528
 529             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 530
 531             /* Update vectorial force */
 532             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 533             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 534             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 535
 536             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 537             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 538             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 539
 540             /**************************
 541              * CALCULATE INTERACTIONS *
 542              **************************/
 543
 544             /* Compute parameters for interactions between i and j atoms */
 545             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 546
 547             /* COULOMB ELECTROSTATICS */
 548             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 549             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 550
 551             /* Update potential sum for this i atom from the interaction with this j atom. */
 552             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 553             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 554
 555             fscal            = felec;
 556
 557             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 558
 559             /* Update vectorial force */
 560             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 561             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 562             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 563
 564             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 565             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 566             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 567
 568             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 569
 570             /* Inner loop uses 155 flops */
 571         }
 572
 573         /* End of innermost loop */
 574
 575         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 576                                               f+i_coord_offset,fshift+i_shift_offset);
 577
 578         ggid                        = gid[iidx];
 579         /* Update potential energies */
 580         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 581         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 582
 583         /* Increment number of inner iterations */
 584         inneriter                  += j_index_end - j_index_start;
 585
 586         /* Outer loop uses 26 flops */
 587     }
 588
 589     /* Increment number of outer iterations */
 590     outeriter        += nri;
 591
 592     /* Update outer/inner flops */
 593
 594     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*155);
 595 }
 596 /*
 597  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 598  * Electrostatics interaction: Coulomb
 599  * VdW interaction:            CubicSplineTable
 600  * Geometry:                   Water4-Particle
 601  * Calculate force/pot:        Force
 602  */
 603 void
 604 nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 605                     (t_nblist                    * gmx_restrict       nlist,
 606                      rvec                        * gmx_restrict          xx,
 607                      rvec                        * gmx_restrict          ff,
 608                      t_forcerec                  * gmx_restrict          fr,
 609                      t_mdatoms                   * gmx_restrict     mdatoms,
 610                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 611                      t_nrnb                      * gmx_restrict        nrnb)
 612 {
 613     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 614      * just 0 for non-waters.
 615      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 616      * jnr indices corresponding to data put in the four positions in the SIMD register.
 617      */
 618     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 619     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 620     int              jnrA,jnrB;
 621     int              j_coord_offsetA,j_coord_offsetB;
 622     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 623     real             rcutoff_scalar;
 624     real             *shiftvec,*fshift,*x,*f;
 625     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 626     int              vdwioffset0;
 627     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 628     int              vdwioffset1;
 629     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 630     int              vdwioffset2;
 631     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 632     int              vdwioffset3;
 633     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 634     int              vdwjidx0A,vdwjidx0B;
 635     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 636     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 637     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 638     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 639     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 640     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 641     real             *charge;
 642     int              nvdwtype;
 643     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 644     int              *vdwtype;
 645     real             *vdwparam;
 646     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 647     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 648     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 649     real             *vftab;
 650     _fjsp_v2r8       itab_tmp;
 651     _fjsp_v2r8       dummy_mask,cutoff_mask;
 652     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 653     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 654     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 655
 656     x                = xx[0];
 657     f                = ff[0];
 658
 659     nri              = nlist->nri;
 660     iinr             = nlist->iinr;
 661     jindex           = nlist->jindex;
 662     jjnr             = nlist->jjnr;
 663     shiftidx         = nlist->shift;
 664     gid              = nlist->gid;
 665     shiftvec         = fr->shift_vec[0];
 666     fshift           = fr->fshift[0];
 667     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 668     charge           = mdatoms->chargeA;
 669     nvdwtype         = fr->ntype;
 670     vdwparam         = fr->nbfp;
 671     vdwtype          = mdatoms->typeA;
 672
 673     vftab            = kernel_data->table_vdw->data;
 674     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 675
 676     /* Setup water-specific parameters */
 677     inr              = nlist->iinr[0];
 678     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 679     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 680     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 681     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 682
 683     /* Avoid stupid compiler warnings */
 684     jnrA = jnrB = 0;
 685     j_coord_offsetA = 0;
 686     j_coord_offsetB = 0;
 687
 688     outeriter        = 0;
 689     inneriter        = 0;
 690
 691     /* Start outer loop over neighborlists */
 692     for(iidx=0; iidx<nri; iidx++)
 693     {
 694         /* Load shift vector for this list */
 695         i_shift_offset   = DIM*shiftidx[iidx];
 696
 697         /* Load limits for loop over neighbors */
 698         j_index_start    = jindex[iidx];
 699         j_index_end      = jindex[iidx+1];
 700
 701         /* Get outer coordinate index */
 702         inr              = iinr[iidx];
 703         i_coord_offset   = DIM*inr;
 704
 705         /* Load i particle coords and add shift vector */
 706         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 707                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 708
 709         fix0             = _fjsp_setzero_v2r8();
 710         fiy0             = _fjsp_setzero_v2r8();
 711         fiz0             = _fjsp_setzero_v2r8();
 712         fix1             = _fjsp_setzero_v2r8();
 713         fiy1             = _fjsp_setzero_v2r8();
 714         fiz1             = _fjsp_setzero_v2r8();
 715         fix2             = _fjsp_setzero_v2r8();
 716         fiy2             = _fjsp_setzero_v2r8();
 717         fiz2             = _fjsp_setzero_v2r8();
 718         fix3             = _fjsp_setzero_v2r8();
 719         fiy3             = _fjsp_setzero_v2r8();
 720         fiz3             = _fjsp_setzero_v2r8();
 721
 722         /* Start inner kernel loop */
 723         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 724         {
 725
 726             /* Get j neighbor index, and coordinate index */
 727             jnrA             = jjnr[jidx];
 728             jnrB             = jjnr[jidx+1];
 729             j_coord_offsetA  = DIM*jnrA;
 730             j_coord_offsetB  = DIM*jnrB;
 731
 732             /* load j atom coordinates */
 733             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 734                                               &jx0,&jy0,&jz0);
 735
 736             /* Calculate displacement vector */
 737             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 738             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 739             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 740             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 741             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 742             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 743             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 744             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 745             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 746             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 747             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 748             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 749
 750             /* Calculate squared distance and things based on it */
 751             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 752             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 753             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 754             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 755
 756             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 757             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 758             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 759             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 760
 761             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 762             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 763             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 764
 765             /* Load parameters for j particles */
 766             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 767             vdwjidx0A        = 2*vdwtype[jnrA+0];
 768             vdwjidx0B        = 2*vdwtype[jnrB+0];
 769
 770             fjx0             = _fjsp_setzero_v2r8();
 771             fjy0             = _fjsp_setzero_v2r8();
 772             fjz0             = _fjsp_setzero_v2r8();
 773
 774             /**************************
 775              * CALCULATE INTERACTIONS *
 776              **************************/
 777
 778             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 779
 780             /* Compute parameters for interactions between i and j atoms */
 781             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 782                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 783
 784             /* Calculate table index by multiplying r with table scale and truncate to integer */
 785             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 786             itab_tmp         = _fjsp_dtox_v2r8(rt);
 787             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 788             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 789             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 790
 791             vfconv.i[0]     *= 8;
 792             vfconv.i[1]     *= 8;
 793
 794             /* CUBIC SPLINE TABLE DISPERSION */
 795             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 796             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 797             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 798             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 799             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 800             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 801             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 802             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 803             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 804
 805             /* CUBIC SPLINE TABLE REPULSION */
 806             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 807             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 808             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 809             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 810             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 811             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 812             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 813             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 814             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 815             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 816
 817             fscal            = fvdw;
 818
 819             /* Update vectorial force */
 820             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 821             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 822             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 823
 824             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 825             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 826             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 827
 828             /**************************
 829              * CALCULATE INTERACTIONS *
 830              **************************/
 831
 832             /* Compute parameters for interactions between i and j atoms */
 833             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 834
 835             /* COULOMB ELECTROSTATICS */
 836             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 837             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 838
 839             fscal            = felec;
 840
 841             /* Update vectorial force */
 842             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 843             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 844             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 845
 846             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 847             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 848             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 849
 850             /**************************
 851              * CALCULATE INTERACTIONS *
 852              **************************/
 853
 854             /* Compute parameters for interactions between i and j atoms */
 855             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 856
 857             /* COULOMB ELECTROSTATICS */
 858             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 859             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 860
 861             fscal            = felec;
 862
 863             /* Update vectorial force */
 864             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 865             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 866             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 867
 868             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 869             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 870             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 871
 872             /**************************
 873              * CALCULATE INTERACTIONS *
 874              **************************/
 875
 876             /* Compute parameters for interactions between i and j atoms */
 877             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 878
 879             /* COULOMB ELECTROSTATICS */
 880             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 881             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 882
 883             fscal            = felec;
 884
 885             /* Update vectorial force */
 886             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 887             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 888             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 889
 890             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 891             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 892             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 893
 894             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 895
 896             /* Inner loop uses 144 flops */
 897         }
 898
 899         if(jidx<j_index_end)
 900         {
 901
 902             jnrA             = jjnr[jidx];
 903             j_coord_offsetA  = DIM*jnrA;
 904
 905             /* load j atom coordinates */
 906             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 907                                               &jx0,&jy0,&jz0);
 908
 909             /* Calculate displacement vector */
 910             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 911             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 912             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 913             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 914             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 915             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 916             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 917             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 918             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 919             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 920             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 921             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 922
 923             /* Calculate squared distance and things based on it */
 924             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 925             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 926             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 927             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 928
 929             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 930             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 931             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 932             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 933
 934             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 935             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 936             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 937
 938             /* Load parameters for j particles */
 939             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 940             vdwjidx0A        = 2*vdwtype[jnrA+0];
 941
 942             fjx0             = _fjsp_setzero_v2r8();
 943             fjy0             = _fjsp_setzero_v2r8();
 944             fjz0             = _fjsp_setzero_v2r8();
 945
 946             /**************************
 947              * CALCULATE INTERACTIONS *
 948              **************************/
 949
 950             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 951
 952             /* Compute parameters for interactions between i and j atoms */
 953             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 954
 955             /* Calculate table index by multiplying r with table scale and truncate to integer */
 956             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 957             itab_tmp         = _fjsp_dtox_v2r8(rt);
 958             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 959             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 960             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 961
 962             vfconv.i[0]     *= 8;
 963             vfconv.i[1]     *= 8;
 964
 965             /* CUBIC SPLINE TABLE DISPERSION */
 966             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 967             F                = _fjsp_setzero_v2r8();
 968             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 969             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 970             H                = _fjsp_setzero_v2r8();
 971             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 972             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 973             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 974             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 975
 976             /* CUBIC SPLINE TABLE REPULSION */
 977             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 978             F                = _fjsp_setzero_v2r8();
 979             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 980             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 981             H                = _fjsp_setzero_v2r8();
 982             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 983             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 984             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 985             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 986             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 987
 988             fscal            = fvdw;
 989
 990             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 991
 992             /* Update vectorial force */
 993             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 994             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 995             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 996
 997             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 998             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 999             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1000
1001             /**************************
1002              * CALCULATE INTERACTIONS *
1003              **************************/
1004
1005             /* Compute parameters for interactions between i and j atoms */
1006             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1007
1008             /* COULOMB ELECTROSTATICS */
1009             velec            = _fjsp_mul_v2r8(qq10,rinv10);
1010             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
1011
1012             fscal            = felec;
1013
1014             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1015
1016             /* Update vectorial force */
1017             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1018             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1019             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1020
1021             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1022             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1023             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1024
1025             /**************************
1026              * CALCULATE INTERACTIONS *
1027              **************************/
1028
1029             /* Compute parameters for interactions between i and j atoms */
1030             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1031
1032             /* COULOMB ELECTROSTATICS */
1033             velec            = _fjsp_mul_v2r8(qq20,rinv20);
1034             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
1035
1036             fscal            = felec;
1037
1038             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1039
1040             /* Update vectorial force */
1041             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1042             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1043             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1044
1045             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1046             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1047             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1048
1049             /**************************
1050              * CALCULATE INTERACTIONS *
1051              **************************/
1052
1053             /* Compute parameters for interactions between i and j atoms */
1054             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1055
1056             /* COULOMB ELECTROSTATICS */
1057             velec            = _fjsp_mul_v2r8(qq30,rinv30);
1058             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
1059
1060             fscal            = felec;
1061
1062             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1063
1064             /* Update vectorial force */
1065             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1066             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1067             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1068
1069             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1070             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1071             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1072
1073             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1074
1075             /* Inner loop uses 144 flops */
1076         }
1077
1078         /* End of innermost loop */
1079
1080         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1081                                               f+i_coord_offset,fshift+i_shift_offset);
1082
1083         /* Increment number of inner iterations */
1084         inneriter                  += j_index_end - j_index_start;
1085
1086         /* Outer loop uses 24 flops */
1087     }
1088
1089     /* Increment number of outer iterations */
1090     outeriter        += nri;
1091
1092     /* Update outer/inner flops */
1093
1094     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*144);
1095 }